Posted by alexandra_k on January 26, 2014, at 2:57:34
In reply to Corelation does not imply causation, posted by poser938 on January 25, 2014, at 22:11:09
correlation not implying causation is part of the rationale behind randomized double blind control trials being considered best standard of evidence.
because individuals are randomly assigned to treatment or control group (and nobody knows which group anybody is in) the ONLY difference between the control group and the treatment group (with respect to the populations) is the presence of the medication vs placebo.
thus if you find population level differences between one group feeling better or getting better or whatever for your measure of better... one can infer that the cause of that was the treatment.
without the control group - one would be left making (problematic) inferences from correlation to causation.
of course it might be that even though the people were randomly assigned it just happens to be the case that the people taking the active meds have dogs and the people in the control group don't. or the people in the active group are all under 25 but the people in the control group are over 40. or whatever... there are guidelines on sample sizes and margins of error to try and reduce these sorts of possible confounds...
though...
to be fair...
drug trials isn't so much science... as it is about the process of patenting an invention / product.
often it isn't about efficacy so much as about less side-effects than previous generation medication. efficacy... never was particularly good, i don't think. i'm fairly sure nobody much is claiming that newer generation medications are *more effective*. i was pretty sure *safer* was their main claim to fame... (largely because longitudinal data ain't in yet). which is convenient... given the life of a patient...
poster:alexandra_k
thread:1059520
URL: http://www.dr-bob.org/babble/20140123/msgs/1059533.html